Land cover change - mapping forest extent, clearing and regrowth
- Overview
- Implementation
- Integration and analysis
- Technical reports
- Time sequence of land cover change
Overview
The National Carbon Accounting System operates one of the largest satellite monitoring programs in the world, providing a comprehensive picture of land cover change over the Australian continent for the past 30 years. The area covers around 7.5 million square kilometres, with environments ranging from desert to rainforest, and through a period of accelerated landscape change. The NCAS uses Landsat satellite imagery to map forest extent at 25 m resolution for several time periods since 1972. A forest is defined as vegetation with a minimum 20 per cent canopy cover, potentially reaching 2 metres high and a minimum area of 0.2 hectares.
All images are geographically and spectrally calibrated to a standard base. Advanced statistical techniques compare and validate changes in individual pixels, and eliminate false change records due to climatic or other environmental effects. Maps of forest cover, clearing and regrowth are produced for each time period. These are attributed for cause of change, as the greenhouse account excludes change that is not attributable to direct human actions.
Field data, aerial photographs and high resolution satellite imagery are used for verifying the change analysis.

Implementation
Implementation of a standardised, high quality national land cover change analysis involved several key steps:
- development of a nationally agreed specification through consultation - covering image calibration, processing and analysis
- development of software and baseline data to calibrate images to a common base and analyse for change
- refinement of techniques through pilot projects
- selection of suitable satellite images from 1972 onwards
- subsequent rectification, spectral calibration and analysis of all images.
Landsat MSS, TM and ETM+ data for 12 time periods from 1972 to 2002 are re-sampled to 25 m pixels and calibrated to a standard geographic and spectral base, developed from the Year 2000 coverage of Landsat ETM+ data. Mapping of forest extent is automated through a process known as thresholding, that uses a set of reference indices to identify forest in select zones. Presence or absence of forest is determined for each 25 m pixel at each time period, by comparing the spectral index in the Landsat image with the reference index for that zone. The reference indices were prepared using aerial photographs and skilled interpreters.
Maps of clearing and revegetation are prepared by comparing forest pixels through subsequent time periods. A Conditional Probability Network (CPN) analysis and subsequent consistency check ensures that the pixel classification (either forest or non-forest) is logically consistent through space and time.
The final step is to attribute each of the 'change' pixels (clearing or re-vegetation) for cause of change. Change which is not attributable to direct human action is masked out, as it is not part of the greenhouse account. The attribution maps are prepared separately through a range of sources and include fire, agricultural activity, conservation areas and areas managed for forestry.

Implementation has involved several commercial service providers for each step of the analysis, all working within a standard quality assurance framework. Pilot tests were used in the initial stages to train/develop industry capacity in Australia, refine methods and software, and develop logistical systems to maximise both output and opportunity for progressive quality assurance/quality control.
Analytical methods are still being refined through a continuous improvement program aimed at reducing uncertainties. Analytical changes are incorporated as further satellite images and time intervals are added to the sequence of national coverages.
Integration and analysis
Land cover change data are used within the FullCAM modelling framework to provide a 30-year dynamic perspective on the nature and extent of change in land systems across the continent over the period since 1970.
- Analytical methods and supplement (these link directly to bookmarks within a pdf document)
- FullCAM model
- Latest results
Technical reports
- Update on the National Carbon Accounting System Continuous Improvement and Verification Methodology - TR 46, 2004
- Continuous Improvement of the National Carbon Accounting System Land Cover Change Mapping - TR 39, 2003
- Satellite Monitoring of Land-Based Sources and Sinks of Greenhouse Gases - Fact sheet, July 2003
- Pilot Testing Of Remote Sensing Methodology For Mapping Land Cover Change - TR 16, 2002
- Land Cover Change: Specification for Remote Sensing Analysis - TR 9, 2002
- Land Clearing: A Social History - TR 4, 2001

